The Genetics of Myeloid Malignancies: from Germline Risk to Somatic Transformation

Transcript

• 00:01So it's my pleasure to
• 00:03welcome,
• 00:04doctor Coleman Lindsley,
• 00:07our invited speaker for today's
• 00:09Blanche Toulmin lecture series and
• 00:12Yale Cancer Center Grand Rounds.
• 00:14So the Blanche Toulmin lecture
• 00:16series, was established in two
• 00:18thousand twelve
• 00:20by doctor Marvin Sears.
• 00:22Doctor Sears was a long
• 00:24time chair and founder of
• 00:25ophthalmology
• 00:26and visual science at Yale,
• 00:28and the lecture was established
• 00:30in honor of his mother,
• 00:31Blanche Tollman,
• 00:33who passed away from AML.
• 00:36This was the first lecture
• 00:38series at Yale dedicated solely
• 00:39to hematologic malignancies,
• 00:42and it is intended to
• 00:43bring to Yale pioneers,
• 00:46like doctor Lindsley,
• 00:47that have made major contributions
• 00:49to our understanding
• 00:51of the current trends in
• 00:53hematologic
• 00:54oncology.
• 00:57So a little bit more,
• 00:58on doctor Lindsley. So doctor
• 01:01Lindsley is an associate professor
• 01:03of medicine
• 01:04at Harvard Medical School,
• 01:06and at the Dana Farber
• 01:07Cancer Institute,
• 01:09where he is also the
• 01:10director of the Edward p
• 01:12Evans Center or MDS.
• 01:15He received his MD PhD
• 01:17in immunology,
• 01:18from Washington University School of
• 01:21Medicine. He did his residency
• 01:22at the Brigham and Women's
• 01:24Hospital. He did his fellowship
• 01:26at the Dana Farber.
• 01:28And amongst his,
• 01:31wonderful activities and leadership in
• 01:33the field of MDS,
• 01:35he's a member of the
• 01:36MDS genetic subcommittee for the
• 01:38NIH national MDS study.
• 01:41He is on the steering
• 01:41committee for the NHLBI
• 01:43transomics
• 01:44for precision medicine,
• 01:46group, and he is on
• 01:47the molecular committee for the
• 01:49international working group for prognosis,
• 01:52in MDS.
• 01:54His laboratory,
• 01:55at the Dana Farber,
• 01:58focuses on the biology and
• 01:59the treatment of myeloid malignancies.
• 02:02His genetic studies
• 02:04have
• 02:05changed,
• 02:06and created new paradigms in
• 02:08our field,
• 02:09specifically new genomic models of
• 02:11leukemia classification,
• 02:13and
• 02:14of
• 02:15how to understand MDS outcomes
• 02:17after allogeneic stem cell transplant.
• 02:20His laboratory uses a variety
• 02:22of approaches, including mouse and
• 02:23cell line models to dissect
• 02:25the mechanisms
• 02:28behind genetic cooperation
• 02:30during the progression of myeloid
• 02:32diseases,
• 02:33And he has a very
• 02:35interesting and specific focus on
• 02:37leukemia initiation within the context,
• 02:40of predisposition
• 02:41syndromes,
• 02:43to,
• 02:44myeloid malignancies
• 02:45as well as studying mutations,
• 02:48in the field of epigenetics.
• 02:50So
• 02:51before starting, I think
• 02:53there is a token that
• 02:55we would like
• 02:57to give,
• 02:58so that you can remember
• 03:00your time with us. Thank
• 03:02you. Welcome.
• 03:14Thanks very much for the
• 03:15introduction.
• 03:16It's too kind,
• 03:18and, for the invitation.
• 03:21It's been a pleasure to
• 03:22to see some,
• 03:24old friends and meet some
• 03:25new ones, including some of
• 03:27the,
• 03:29rising stars in our field,
• 03:31just before lunch.
• 03:33And so,
• 03:35I'll dive in,
• 03:37to talk about,
• 03:40about myeloid genetics.
• 03:41Here's some disclosures. None of
• 03:43them are relevant for what
• 03:44I'm gonna talk about today.
• 03:47I think if I were
• 03:48to,
• 03:49simplify down
• 03:51what I'm interested in,
• 03:53It's essentially the origin story
• 03:56of
• 03:56myeloid leukemias,
• 03:59you know, or you can
• 03:59call it the ontogeny
• 04:01of of disease.
• 04:03And this is essentially how
• 04:04we get from a normal
• 04:05stem cell,
• 04:07that functions
• 04:08properly to a florid myeloid
• 04:10malignancy,
• 04:12with failure of the bone
• 04:14marrow organ.
• 04:15And the origin,
• 04:17commonly is just aging related
• 04:19degeneration
• 04:21of the stem cell pool.
• 04:23But we can also have
• 04:23acquired predispositions,
• 04:27as well as exposures to
• 04:29leukemogenic agents like chemotherapy and
• 04:31radiation,
• 04:32as well as an increasingly
• 04:35recognized number of inherited predispositions
• 04:37that modify
• 04:38the probability of of,
• 04:40developing,
• 04:42clonal myeloid disease in life.
• 04:44And so it's
• 04:46this origin,
• 04:48that I seek to understand,
• 04:51and more importantly, try to
• 04:52understand how that origin influences
• 04:55the outcome of patients,
• 04:56the progression of patients, the
• 04:58response to therapies,
• 05:00and,
• 05:01offers us opportunities for, development
• 05:04of novel, therapeutic approaches.
• 05:06And so to bring us
• 05:07back to the to the
• 05:08basics,
• 05:09the cell of,
• 05:11origin here is the hematopoietic
• 05:13stem cell, which functions normally
• 05:15with its dual roles of,
• 05:17I wouldn't say perpetual self
• 05:19renewal,
• 05:20but ongoing self renewal throughout
• 05:21life and its capacity for
• 05:23multilinear differentiation
• 05:25into the cells of the
• 05:26blood, to carry oxygen,
• 05:29fight infections,
• 05:30clot the blood.
• 05:32And rarely,
• 05:34or maybe not so rarely,
• 05:36one of these stem cells
• 05:37acquires a mutation
• 05:38at some point in life
• 05:40and
• 05:41grows,
• 05:42at a faster rate than
• 05:43its normal neighbors.
• 05:45And
• 05:46at its base basic level,
• 05:47this is clonal hematopoiesis.
• 05:50And in that first step,
• 05:51there's rarely any detectable impact
• 05:53on the function of the
• 05:54organ.
• 05:56The cells are apparently normal
• 05:58in number,
• 05:59and
• 06:01and there's
• 06:03limited effect,
• 06:04at the individual level.
• 06:06Clonometopoiesis
• 06:07in certain genes like DNMT
• 06:09three a have been linked
• 06:10and tattoo have been linked
• 06:11with,
• 06:12adverse cardiovascular outcomes, immune inflammatory
• 06:14signaling,
• 06:17and have a whole host
• 06:18of,
• 06:19immune
• 06:20derangements that drive
• 06:22non malignant pathology.
• 06:24And then,
• 06:25on the other side, clones
• 06:27can,
• 06:28acquire,
• 06:29a sequential,
• 06:32subclonal progression via additional
• 06:34gene mutations to develop a
• 06:36frank leukemia.
• 06:37And this leukemia,
• 06:42is associated with organ failure
• 06:44or cytopenias.
• 06:45And so it's really this
• 06:47process that, that we're gonna
• 06:49focus on today. There's been
• 06:51a lot of study over
• 06:52the past now fifteen years
• 06:54or more about
• 06:55the spectrum of recurrently mutated
• 06:57genes
• 06:58in the development of myeloid
• 07:00malignancies.
• 07:01And these range from those
• 07:03that alter DNA methylation,
• 07:05chromatin organization,
• 07:06RNA splicing, DNA damage response,
• 07:08growth factor receptor signaling,
• 07:11myeloid transcription factors.
• 07:13They all share the common
• 07:14property of in the right
• 07:15context.
• 07:17They,
• 07:18drive clonal
• 07:19advantage or clonal dominance.
• 07:22But it's important to recognize
• 07:24that they don't all do
• 07:25this
• 07:26in the same way,
• 07:28or in the same,
• 07:30combination or in
• 07:32random order. There's actually a
• 07:34highly stereotyped
• 07:35progression
• 07:37from initiation
• 07:39through,
• 07:40intermediate
• 07:41stage of progression
• 07:43through to terminal transformation to
• 07:45a fluid leukemia. And there
• 07:46are specific genes that are
• 07:48associated with each of these
• 07:49steps.
• 07:50And for orientation here, there's
• 07:52this is a very simple
• 07:54productionist model of the kind
• 07:56of, three large categories of
• 07:59myeloid leukemia,
• 08:01those that have,
• 08:03MDS associated
• 08:04mutations,
• 08:07like those impacting RNA splicing
• 08:09and chromatin modification
• 08:10and one of the cohesins.
• 08:13P fifty three,
• 08:14defines
• 08:15in some cases its own,
• 08:18distinct
• 08:19ontogeny or type of disease.
• 08:21And then there's de novo
• 08:22AML, which,
• 08:24is a little less degenerative
• 08:26in its origin and then
• 08:27and is, more genetically simple.
• 08:31And,
• 08:32the reason I present a
• 08:34reductionist model is because it's
• 08:36step one in trying to
• 08:37understand,
• 08:39the how these diseases progress,
• 08:41how they respond to treatment.
• 08:44And I'll give you one
• 08:45example of that.
• 08:48At first blush,
• 08:49this model was simply classification.
• 08:52And
• 08:53the,
• 08:54some of its features have
• 08:55been incorporated into,
• 08:57classification and prognostic models over
• 08:59the years,
• 09:01and,
• 09:03allow us to,
• 09:05reframe or reclassify,
• 09:08historical
• 09:08clinical trials, for example. And
• 09:10I'm just gonna show you
• 09:11one, one example.
• 09:13And so in two thousand
• 09:15thirteen or fourteen,
• 09:16Cellator and Jazz,
• 09:19planned the phase three trial,
• 09:22comparing,
• 09:23CPX three five one, a
• 09:25liposomal preparation of dongorubicin, cytarabine
• 09:27to standard
• 09:29infusional seven plus three for
• 09:31high risk patients with secondary
• 09:33and therapy related AML.
• 09:36This was based on clinical,
• 09:39inclusion criteria defining high risk,
• 09:41and this
• 09:42was AML with myelodysplasia related
• 09:44changes,
• 09:46defined by
• 09:47recurrent cytogenetic abnormalities
• 09:50or a history of
• 09:53clinical MDS or CMML.
• 09:55And then the third group
• 09:57was those who had had
• 09:58a,
• 09:59chemotherapy or cytotoxic therapy exposure.
• 10:01And this was according to
• 10:02the two thousand eight WHO,
• 10:06criteria for for this high
• 10:07risk group of patients. And
• 10:09as you can see on
• 10:09the right, CPX three five
• 10:11one, when group when analyzed
• 10:13in all these patients was
• 10:14associated with an improved
• 10:16overall survival. And on this
• 10:17basis,
• 10:19the drug received FDA approval.
• 10:22In
• 10:23great shock and disappointment to
• 10:25the company,
• 10:26the approval indication,
• 10:29evaporated,
• 10:30shortly after they received it,
• 10:32meaning we started reclassifying
• 10:34disease,
• 10:36using different terms. And so,
• 10:39now, they're,
• 10:42the notion of biological,
• 10:45disease or ontogeny,
• 10:48is,
• 10:49a greater explanatory model for
• 10:51disease behavior than the clinical
• 10:53observation. So,
• 10:55these groups can be,
• 10:57these clinical groups can be
• 10:58reclassified
• 10:59based on genetics into some
• 11:01some categories I'm showing you
• 11:02here,
• 11:03according to
• 11:06WHO and ICC classification
• 11:08as well as ELN,
• 11:11prognostic models.
• 11:12And so what does this
• 11:13trial look like when when
• 11:15reevaluated?
• 11:16Here are the three main
• 11:17inclusion
• 11:18groups.
• 11:20On the left AML MRC
• 11:21with prior MDS or CMML,
• 11:24so transforming after a chronic
• 11:26myeloid malignancy.
• 11:28In the middle,
• 11:30basically
• 11:31complex karyotype chromosomes five, seven,
• 11:33seventeen abnormalities, and on the
• 11:35right, post cytotoxic therapy. And
• 11:37we can see is that
• 11:38these are very heterogeneous groups
• 11:40when we look at,
• 11:41genetics.
• 11:43Those with prior MDS or
• 11:45CMML were largely
• 11:46had MDS associated mutations,
• 11:50with about twenty percent with
• 11:52p fifty three mutations.
• 11:53Fifteen percent or so looked
• 11:55like they had de novo
• 11:56AML.
• 11:57The AML MRC was cytogenetics
• 11:59and therapy related were each
• 12:01more than fifty percent p
• 12:02fifty three,
• 12:04and then half, something else.
• 12:07And so when we now
• 12:09reshuffle the deck,
• 12:11to have more genetically concordant
• 12:13groups,
• 12:14this is how the outcomes
• 12:15look when we compare clinical
• 12:17classification
• 12:18and genetic classification from this
• 12:20phase three trial. On the
• 12:21left, these three clinical groups
• 12:23overlay their Kilometers curves over
• 12:25overlap each other. When we
• 12:27use these four genetic groups,
• 12:29and I'm adding DDX forty
• 12:30one as a, recurrent germline
• 12:32contributor,
• 12:34we see a widespread even
• 12:35in this extremely high risk
• 12:37group of patients.
• 12:39And so
• 12:41the genetic classification affords opportunity
• 12:43to better resolve outcomes in
• 12:45this group.
• 12:47So how does the drug
• 12:48look? Does the drug work
• 12:49differently in different
• 12:51types of disease?
• 12:53Here are the four categories
• 12:54now broken down by treatment
• 12:56arm. You can see in
• 12:58the bottom left p fifty
• 12:59three, no difference.
• 13:01And in fact, the bulk
• 13:01of the signal is within
• 13:03the group of patients who
• 13:04have AMLMR mutations.
• 13:06Those who have de novo
• 13:07disease, no difference.
• 13:09The numbers are really small
• 13:10here in the DDX group.
• 13:12There's a hint that maybe
• 13:13they respond better, but I'm
• 13:14gonna leave that for future
• 13:15studies. But really quantitatively,
• 13:18this group here, AMLMR,
• 13:20drive
• 13:21drives and drove the clinical
• 13:23effect that was seen in
• 13:24the fan in the randomized
• 13:25trial.
• 13:27Even more so,
• 13:29what drives that benefit?
• 13:31It's really the ability to
• 13:32get to transplant.
• 13:33And so here I'm showing
• 13:35you on the left,
• 13:36no transplant seven and three
• 13:38versus CPX. The curves overlap,
• 13:40and they're steeply down. And
• 13:42then here is seven plus
• 13:43three versus CPX among those
• 13:45who got to transplant.
• 13:46And so something about it,
• 13:48they didn't include the adequate
• 13:49correlatives to really explain why.
• 13:53But AMLMR
• 13:54who go to transplant
• 13:56actually have a shockingly favorable
• 13:58prognosis
• 13:59here for this group who
• 14:01in the absence of transplant
• 14:02actually are largely,
• 14:05dead by one year.
• 14:07And this is this holds
• 14:08true in a in a
• 14:09extensive multivariable model
• 14:11here. So treatment and transplant
• 14:14are both independently predictive of
• 14:16of outcome,
• 14:17among patients with AML MR.
• 14:20So
• 14:22just
• 14:24p fifty three. I have
• 14:25to speak to it very
• 14:26briefly. We looked at this
• 14:27group and said,
• 14:29is there any treatment effect?
• 14:30No. There wasn't.
• 14:32But we also re we
• 14:34classify these based on,
• 14:36empirical determination of allelic state,
• 14:40by looking at, NGS determination
• 14:42of copy neutral,
• 14:44loss of heterozygosity or deletion
• 14:46of seventeen p in the
• 14:47presence of a mutation,
• 14:49or, karyotypic
• 14:51chromosome seventeen alteration
• 14:53or the presence of two
• 14:54mutations,
• 14:55two point mutations and categorize
• 14:57them as p fifty three
• 14:58single or p fifty three
• 14:59multi hit.
• 15:01What you can see here
• 15:02is most were multi hit
• 15:03as is common in AML.
• 15:05Most of them were pure
• 15:07p fifty three disease, meaning
• 15:08they look like they started
• 15:10as p fifty three. They
• 15:11progressed as p fifty three,
• 15:12and that's true p fifty
• 15:14three ontogeny.
• 15:15There's this other small group
• 15:17and most of the signal
• 15:18singles, which have
• 15:20subclonal progression
• 15:22of their MDS with a
• 15:23p fifty three mutation.
• 15:25And,
• 15:26and this,
• 15:27in work I won't talk
• 15:28to you here about is,
• 15:30these two types of p
• 15:31fifty three to find not
• 15:32only by the state, but
• 15:34by genetic context really drives
• 15:36clinical outcome. And here, it's
• 15:38really the lolic state and
• 15:39not the treatment. This is
• 15:40just looking at lolic state.
• 15:41So p fifty three single
• 15:42in this group of patients
• 15:43overlays with p fifty three
• 15:45absent,
• 15:46and multi is what drives
• 15:47the adverse effect. And in
• 15:49and there was no,
• 15:52treatment effect for the among
• 15:54the multis.
• 15:55So allelic state in p
• 15:57fifty three induction outcomes.
• 15:59We also looked at here
• 16:00is because if you think
• 16:01about how do you wanna
• 16:02apply a new therapy to
• 16:04the patient who's in front
• 16:04of you, you need to
• 16:05know its effect, and you
• 16:07need to know its consequence
• 16:09to the patient outside of
• 16:10its on target effect. And
• 16:12so we looked here,
• 16:14on the left is, among
• 16:15patients who got into CR,
• 16:17their time to count recovery
• 16:18to an ANC of five
• 16:19hundred. On the right, count
• 16:20recovery to a platelet count
• 16:21of fifty, among those who
• 16:23had CR. And here's just
• 16:24days from induction with a
• 16:26with a dotted
• 16:27line as twenty eight days.
• 16:30And here I'm showing you
• 16:31each genetic group,
• 16:35paired with their seven plus
• 16:36three on top and CPX
• 16:38next to it.
• 16:40We can see for both
• 16:41ANC and platelet recovery,
• 16:44CPX,
• 16:44independent of genetic subgroup, was
• 16:46associated with longer time to
• 16:48count recovery.
• 16:50But that was particularly evident
• 16:52in,
• 16:53the AMLMR
• 16:54patients who had poorly functioning
• 16:56marrow.
• 16:57And also, I think importantly,
• 17:01extended in these DDX
• 17:02forty one patients,
• 17:05and then these de novo,
• 17:08patients down here.
• 17:11And
• 17:13the last thing I'll show
• 17:13you about this trial is
• 17:14now if we use these
• 17:16genetic subgroups to now define
• 17:18early mortality,
• 17:21it's,
• 17:22sobering,
• 17:23to me.
• 17:25And so DDX forty one,
• 17:27they all survived the first
• 17:28sixty days.
• 17:30And then in,
• 17:32the de novo AML and
• 17:33AMLMR patients, they were about,
• 17:36fifteen to eighteen percent,
• 17:38sixty day mortality.
• 17:40And then the p fifty
• 17:41threes, the multis had a
• 17:43greater than twenty five percent
• 17:45sixty day mortality.
• 17:47And so if you,
• 17:49we can kind of cogitate
• 17:51on that, for a moment.
• 17:52So high risk group of
• 17:53patients,
• 17:56and a lot of early
• 17:57mortality even among,
• 17:59the different
• 18:00groups.
• 18:01And so conclusions from this
• 18:03is when we reclassify,
• 18:06clinical history
• 18:07using biologically
• 18:09driven,
• 18:10genetic subgroups,
• 18:12or the other way, genetic
• 18:14the driven biologic subgroups.
• 18:16You tell me.
• 18:17AMLMR
• 18:18drives the benefit of CPX
• 18:20three five one over seven
• 18:21plus three,
• 18:22and this effect was mediated
• 18:24by,
• 18:25transplant consolidation.
• 18:28For p fifty three,
• 18:30there's no difference in treatment,
• 18:32and allelic status is the
• 18:33main thing.
• 18:34But really, I'd ask you,
• 18:36should these patients get intensive
• 18:38induction? This is a side
• 18:39conversation we can,
• 18:41talk about for a while.
• 18:42More than twenty five percent
• 18:43early mortality,
• 18:45no real benefit.
• 18:48So,
• 18:49I think important to think
• 18:50about. And then the there's
• 18:52delayed count recovery that we
• 18:53should be aware of particularly
• 18:54among the AML or MR
• 18:55patients who achieve CR.
• 18:58So this sort of approach
• 19:00now of reclassification is being
• 19:02applied to
• 19:03Azovent,
• 19:05and
• 19:06novel therapeutics for AML.
• 19:08And,
• 19:09there are different,
• 19:11response
• 19:12characteristics
• 19:13and toxicity characteristics on the
• 19:16these this very simple reductionist
• 19:19categorization of disease.
• 19:22And so,
• 19:24so I just wanted to
• 19:25highlight that as an example.
• 19:26It's kind of like a
• 19:27an old story that is
• 19:29now kind of,
• 19:30coming,
• 19:31forward.
• 19:32I wanna move to some,
• 19:34into some different space, which
• 19:36is,
• 19:38really trying to understand how
• 19:39the germline state
• 19:41modifies the risk of developing
• 19:44myeloid malignancy,
• 19:45but also modifies
• 19:47the path to malignancy.
• 19:49And,
• 19:51and so,
• 19:53what I'll talk about
• 19:57is the initiation of clonal
• 19:59myeloid disease
• 20:00and the concept that somatic
• 20:02clones
• 20:03are
• 20:04selected,
• 20:06by
• 20:07fitness constraints,
• 20:08that are defined
• 20:09by,
• 20:10the germline,
• 20:11that are defined,
• 20:14by aging,
• 20:16and exposures.
• 20:17And so here in the
• 20:18germline encoded fitness, here's a
• 20:20model and then I'll return
• 20:21to this a little bit.
• 20:22In this middle band is
• 20:23normal
• 20:24fitness,
• 20:25and
• 20:27the,
• 20:28clonal dominance is really measured
• 20:30by the fitness of a
• 20:31clone relative to whatever the
• 20:33baseline is for that
• 20:35bone marrow. And so,
• 20:37in
• 20:39marrow failure syndromes or predispositions,
• 20:42many times that baseline fitness
• 20:44is much lower.
• 20:46And so there are many
• 20:48ways,
• 20:49where a somatic mutation can
• 20:51augment the fitness relative to
• 20:53the background,
• 20:56but not all of them
• 20:57drive leukemia.
• 20:58And what I'll describe is
• 21:00the difference between,
• 21:02those paths. And then this
• 21:03is now looking at the,
• 21:07fitness over time, and this
• 21:08happens in everyone.
• 21:10There's a gradual loss of
• 21:11fitness that occurs with aging,
• 21:14and that creates a novel
• 21:16selection pressure with aging or
• 21:18a fitness constraint
• 21:19with aging that that allows
• 21:21for the selection of, of
• 21:23clones.
• 21:24And so the hypothesis that
• 21:25we go into with this
• 21:25is that the age
• 21:27and gene distribution
• 21:29of onomatopoiesis
• 21:31in any given
• 21:33clinical scenario
• 21:35reflects
• 21:35the mechanism
• 21:37and the magnitude
• 21:39of fitness constraint.
• 21:42So that composite of how
• 21:44is the fitness constraint defined
• 21:46and how big is that
• 21:47constraint,
• 21:48or how tight is that
• 21:49constraint
• 21:50really influences
• 21:51the initiation of clonal disease.
• 21:55And so for, this
• 21:58inherited,
• 21:59marrow failure state,
• 22:01baseline fitness is
• 22:03uniformly low,
• 22:04in in the ones I'll
• 22:07talk about.
• 22:08And,
• 22:09we see a lot of
• 22:09clonal metopoiesis. And there are
• 22:11questions that arise.
• 22:12Is this neutral drift because
• 22:15you have a reduced,
• 22:17pool of of of stem
• 22:19cells? So you just have
• 22:21a less complexity and then
• 22:23neutral drift.
• 22:25Are there disease specific factors
• 22:27that drive,
• 22:28selection?
• 22:30Does the path or the
• 22:32the type of CH that
• 22:34or clonometopoiesis
• 22:35that we see, does that
• 22:37determine or drive the leukemia
• 22:39predisposition?
• 22:40And can we use a
• 22:41rational
• 22:42understanding of mechanisms of leukemogenesis
• 22:45and mechanisms of fitness constraint
• 22:47to develop a rational surveillance
• 22:49strategy for patients with ultra
• 22:51high risk of, progression?
• 22:54And can that rational surveillance
• 22:55strategy be the basis for
• 22:57preemptive
• 22:58clinical intervention
• 22:59to mitigate the risks that
• 23:01are associated with transformation?
• 23:04And so here's the paradigm
• 23:05that I'll circle back to.
• 23:07I just wanna put it
• 23:08in your head to start
• 23:09with. And so, again,
• 23:10just to beat the dead
• 23:11horse, fitness constraint fitness baseline
• 23:14fitness is low and uniformly
• 23:16low in many of these
• 23:17marrow failure syndromes.
• 23:19Anything that
• 23:20can normalize that defect
• 23:23will cause a clone,
• 23:26to,
• 23:27grow better than its baseline
• 23:30impaired neighbor.
• 23:32And then there's a subset
• 23:33of mutations which may mediate
• 23:35transformation,
• 23:36through loss of fitness sensing,
• 23:38through bypassing
• 23:40tumor suppression mechanisms.
• 23:42And unlike the normalization where
• 23:44the tumor suppression mechanisms are
• 23:46still intact,
• 23:48on the right, tumor suppression
• 23:49mechanisms are lost and,
• 23:51leukemia ensues.
• 23:53And so how did we
• 23:54go about,
• 23:56evaluating this this,
• 23:58this paradigm?
• 23:59And so this really rose
• 24:01out of a,
• 24:03I would say, serendipitous
• 24:05discovery,
• 24:07which is we looked at
• 24:08a whole bunch of, MDS
• 24:09patients from six months old
• 24:11through
• 24:13seventy five years old, seventy
• 24:14seven,
• 24:15who had allotransplant.
• 24:17And then we looked at
• 24:18those,
• 24:19characteristics that are genetic characteristics
• 24:22that are associated with age,
• 24:25in MDS. And so over
• 24:27here, you see the classic,
• 24:29splicing factor p fifty three,
• 24:30DMT three a tattoo, older
• 24:32age.
• 24:33Younger age,
• 24:34acquired predisposition, Pig a, GATA
• 24:37two,
• 24:38and biallelic
• 24:39SBDS mutations.
• 24:41When we look at outcome
• 24:42among these three groups,
• 24:44though those with Pig a
• 24:45or Gata two, they do
• 24:47really well.
• 24:48Those with SBDS
• 24:50mutations, they do really poorly.
• 24:52And this has been now
• 24:53replicated in other,
• 24:55studies.
• 24:57So SMDS for some reason
• 24:59is associated with very poor
• 25:00outcomes.
• 25:01Now cut to the chase.
• 25:02It's because they all have
• 25:03p fifty three mutations.
• 25:04So that is what drives
• 25:06leukemogenesis, and that's what drives
• 25:08poor outcomes,
• 25:10in leukemia and SDS patients.
• 25:12Most of you may not
• 25:14be familiar with
• 25:16syndrome. It's a disease of,
• 25:18impaired
• 25:19translation,
• 25:21that's mediated by a ribosome
• 25:23maturation defect.
• 25:25And so normally,
• 25:26the,
• 25:28the nascent ribosomal components of
• 25:30sixty s and forty s
• 25:31are assemble are kind of
• 25:33are in the, nucleus.
• 25:35The they're exported
• 25:37to the cytoplasm.
• 25:39Sixty s is,
• 25:41maintained in an unjoined state
• 25:43by binding of of EIF
• 25:45six, which is an anti
• 25:46association factor.
• 25:49SBDS,
• 25:50the gene that's mutated in
• 25:51Schwackman Diamond syndrome,
• 25:54works with EFL one, the
• 25:55GTPase,
• 25:56to kick e e I
• 25:57six off the nascent sixty
• 25:59s, allowing it to join
• 26:01the forty s
• 26:02to create the mature
• 26:04translationally active ADS ribosome.
• 26:06That's how it's normally regulated.
• 26:10Stepwise, ribosome joining,
• 26:13and translation.
• 26:14When you lose
• 26:16SBDS,
• 26:18you lose the ability to
• 26:19kick EIF six off
• 26:21the nascent sixty s, and
• 26:23you get a accumulation of
• 26:25these,
• 26:26free sixty s EIF six
• 26:31molecules in the in the
• 26:32cytoplasm.
• 26:34And you have a reduction
• 26:35in the overall abundance of
• 26:36ADS,
• 26:38translationally active ribosomes,
• 26:41and a downstream
• 26:42reduction in protein translation,
• 26:44which activates p fifty three,
• 26:46which drives bone marrow failure.
• 26:48You can get a sense
• 26:49here. This is gonna be
• 26:50a tight fitness constraint
• 26:52on cellular growth if you
• 26:53can't translate and you're activating
• 26:54p fifty three at baseline.
• 26:56And so that's what happens.
• 26:57These patients have,
• 26:59short stature, exocrine pancreatic insufficiency,
• 27:01and a remarkably high risk
• 27:02of developing,
• 27:04leukemia,
• 27:05oftentimes in the teens and
• 27:07twenties,
• 27:08but now as we're learning,
• 27:09even later.
• 27:10And so to define,
• 27:12the somatic pathways of progression,
• 27:15we did some discovery sequencing
• 27:17in in collaboration with the
• 27:19Schwabman Diamond Syndrome Registry,
• 27:21in patients with and without,
• 27:24leukemia.
• 27:25We identified somatic, recurrent mutations
• 27:27and then used duplex ultrasensitive
• 27:30sequencing
• 27:31in a validation cohort of
• 27:32three hundred twenty seven samples
• 27:34from a hundred and ten
• 27:35patients.
• 27:37Don't ever tell a,
• 27:38pediatric marrow failure patient that
• 27:40that's,
• 27:41doctor that that's not a
• 27:42lot of patients.
• 27:43This is,
• 27:45a huge effort by Akiko
• 27:47Shimomura
• 27:48and the SDS registry to
• 27:49accumulate these serial samples from
• 27:51this from these patients over
• 27:53a long period of time.
• 27:55And so, this is a
• 27:56large group of, SDS patients.
• 27:59All the leukemias had p
• 28:00fifty three mutations.
• 28:02I already said that. Those
• 28:03without leukemia, seventy percent of
• 28:05them had clones.
• 28:06Remember, these are young patients.
• 28:08So how does this compare
• 28:09to regular
• 28:10age associated sporadic CH?
• 28:14Age associated
• 28:15CH,
• 28:16largely DNMT three and tattoo,
• 28:18single mutation,
• 28:19much older age,
• 28:21sixties, seventies, eighties.
• 28:23SDS clonematopoiesis
• 28:25is something totally different.
• 28:27It's ubiquitous by adulthood. So
• 28:29by twenty one here,
• 28:32mind you, there's not that
• 28:33many patients, but they all
• 28:34had,
• 28:35they all had CH. But
• 28:37even in the teenagers, we're
• 28:38talking ninety percent had CH.
• 28:41And even in the first
• 28:42decade of life, more than
• 28:43half had detectable CH.
• 28:45What was that CH?
• 28:47It was EIF six, never
• 28:49been seen to be mutated
• 28:49in humans before,
• 28:51p fifty three, PRPF eight,
• 28:52casein kinase.
• 28:54None of or barely any
• 28:55of these DNMT three a
• 28:56tattoo. So it's not the
• 28:57same CH,
• 28:59and it's not one mutation.
• 29:00Some of these patients had
• 29:01five, ten,
• 29:03fifteen different mutations that are
• 29:05detected, and these are young
• 29:06patients, so it's different. Something's
• 29:08different here.
• 29:09And so this high frequency
• 29:10of EIF six mutations
• 29:12raise the possibility
• 29:14that,
• 29:15maybe,
• 29:17disrupting
• 29:18this,
• 29:20this EI six,
• 29:22RPL twenty three or nascent
• 29:23sixty s interaction,
• 29:25in some ways,
• 29:27promoted
• 29:29ribosome joining and growth of
• 29:30the cell. So here here's
• 29:32what it looks like. ES
• 29:33six is this little tiny
• 29:34cap right here that binds
• 29:37exactly to the site that
• 29:38the four d s binds.
• 29:39And so it's just,
• 29:42inhibits four d s binding
• 29:43through steric hindrance.
• 29:44And when you rotate it
• 29:45out, it's got this beautiful
• 29:48little, like, kinda claw. It's
• 29:50got five fold symmetry with
• 29:51a little, binding area right
• 29:54in the middle
• 29:55here. And so there are
• 29:56potential mechanisms that we imagined.
• 29:59So we could have mutations
• 30:00that disrupt that that binding
• 30:02interaction, that protein protein interaction,
• 30:04or it could just be
• 30:05stoichiometry.
• 30:06You knock out the EF
• 30:07six, you cause haploinsufficiency,
• 30:09and you favor ribosome joining,
• 30:12just by the by the
• 30:14relative concentration
• 30:16of EF six. So there
• 30:18was something like two hundred
• 30:19and fifty different EF six
• 30:20mutations or different,
• 30:23not different mutations themselves, but
• 30:25mutations
• 30:26combined with patients,
• 30:28in this cord. And many
• 30:29of them were missense substitutions.
• 30:31There's those are on top
• 30:33with two very recurrent,
• 30:36alterations here at r ninety
• 30:37six w and m one
• 30:38zero six s,
• 30:39and then also scattered truncating
• 30:42mutations. These were splicing splice
• 30:43site mutations,
• 30:44frame shifts, nonsense,
• 30:46mutations.
• 30:47These miss and substitutions were,
• 30:50almost uniformly in key structural,
• 30:53secondary structure regions as well.
• 30:56And so to kind of
• 30:57imagine or predict what these
• 30:58might do, we developed a
• 30:59homology model of e I
• 31:01six based on a bunch
• 31:03of the published structures in
• 31:04yeast and archaea,
• 31:07and then mapped each mutation
• 31:08onto that homology model and
• 31:10predicted the impact on,
• 31:13on a
• 31:15bunch of characteristics.
• 31:16And so here, that recurrent
• 31:18r ninety six w, here's
• 31:20what it looks like. Normally,
• 31:22the,
• 31:23arginine ninety six and asparagine
• 31:25seventy eight are really in
• 31:26close proximity here in black.
• 31:28These are two hydrogen bonds
• 31:29that kind of keep that
• 31:31together. When you mutate the
• 31:33arginine to a tryptophan,
• 31:35those hydrogen bonds break.
• 31:38The that connection falls apart.
• 31:41And as a result,
• 31:43there's
• 31:44a large change in the
• 31:45free energy resulting in protein
• 31:47destabilization.
• 31:48And so when we re
• 31:49when we express this mutant,
• 31:51at high levels, there's no
• 31:53protein. So it's a it's
• 31:55a
• 31:56protein destabilizing mutation
• 31:58by disrupting these hydrogen bonds.
• 32:00And so
• 32:02is that how all these
• 32:03work?
• 32:04So we map these all,
• 32:05calculated their their,
• 32:07delta delta g's, the free
• 32:08energy changes.
• 32:10And
• 32:11and
• 32:12it's hard to see, but
• 32:14over here on the right
• 32:15is is kind of like
• 32:16the take home message.
• 32:18Over here is a message
• 32:19showing you that we express
• 32:20them, and here is the
• 32:21western blot showing that most
• 32:23of these mutations with high
• 32:24free energy changes
• 32:27are destabilized.
• 32:28And so these are this
• 32:29is just a missense
• 32:31mechanism
• 32:32for resulting in a knockout,
• 32:34of that allele.
• 32:36And
• 32:37we included this one,
• 32:40this as a as a
• 32:41teaser. So n one zero
• 32:42six s, if you remember,
• 32:44was this highly recurrent,
• 32:47mutation.
• 32:48So when we look at
• 32:49that,
• 32:49it's situated right at that
• 32:51interface
• 32:52between
• 32:53e I six and RPL
• 32:54twenty three. And it's these
• 32:57interface mutations here when we
• 32:58look at the at the
• 32:59free energy change.
• 33:01They have minimal impact on
• 33:03protein stability or predicted impact,
• 33:06and all the other missense
• 33:07substitutions have high impact. So
• 33:09here are your destabilizing mutations.
• 33:11And then there's this site
• 33:12which we hypothesize would,
• 33:14not destabilize a protein,
• 33:17but would destabilize the protein
• 33:19protein interaction,
• 33:21destabilize its anti association function,
• 33:23kinda make it fall off.
• 33:25And so these,
• 33:27interface mutations here,
• 33:30have actual increased energy of,
• 33:33binding,
• 33:35as as predicted
• 33:36when we,
• 33:39model that that,
• 33:41RPL twenty three a f
• 33:42six, interface.
• 33:44And so here is just
• 33:45the same thing,
• 33:46where we can show that,
• 33:49this n one zero six
• 33:50s mutation
• 33:51is is,
• 33:53predicted
• 33:53to
• 33:54to, make that, interaction fall
• 33:56apart.
• 33:57And so to really prove
• 33:58it though, we use,
• 34:00sucrose gradient,
• 34:01centrifugation
• 34:02in western blot to really,
• 34:05see whether what the impact
• 34:07of this mutation was on
• 34:08ribosome joining.
• 34:10And so
• 34:12the,
• 34:14the take home here
• 34:15is,
• 34:16usually,
• 34:18there's some some of this
• 34:19free,
• 34:20EF six, but that EF
• 34:22six is usually
• 34:23bound,
• 34:24to the sixty s
• 34:26and then gone from the
• 34:28from the eighty s.
• 34:31When we express
• 34:32the mutant,
• 34:34all of that mutant is
• 34:35in that free fraction.
• 34:37None of it is bound
• 34:38to the sixty s.
• 34:40Only the endogenous here,
• 34:42is bound to the sixty
• 34:43s. And so this,
• 34:45kind of
• 34:46nails the mechanism there that
• 34:48this mutation,
• 34:50simply,
• 34:51breaks apart that that interaction.
• 34:53And,
• 34:54this is supposed to say
• 34:55m one zero six.
• 34:57And what you see is
• 34:58you see increased ADS
• 35:01formation. And so here's the
• 35:03summary from this,
• 35:05from this stuff, which is,
• 35:07the two most frequently mute,
• 35:09mutated genes.
• 35:11ES six mutations
• 35:12repair the ribosome defect.
• 35:14They improve protein translation,
• 35:17and in so doing,
• 35:18decrease p fifty three pathway
• 35:20activation
• 35:21here reflected by CDKM one
• 35:22a expression.
• 35:23P fifty three mutations
• 35:25fail to repair the ribosome
• 35:26defect, fail to improve translation,
• 35:29but obviously still knock out
• 35:30p fifty three function.
• 35:32And so together,
• 35:35they both
• 35:36result in the same end,
• 35:38but one,
• 35:39fixes and one doesn't fix,
• 35:41the underlying issue.
• 35:43So these are frequently found
• 35:44in the same patient.
• 35:46So you could say maybe
• 35:47they are they have a
• 35:48cooperative,
• 35:50function.
• 35:50Maybe they are in classic
• 35:52cancer genetics model where you
• 35:53have one that enables the
• 35:55other one to stick. Or
• 35:56maybe this is just parallel
• 35:58selection in a field of
• 35:59dysfunction.
• 36:00There's different shots on goal
• 36:02of,
• 36:03of clonal outgrowth.
• 36:04And so we did some
• 36:05single cell sequencing in patients
• 36:06with many mutations,
• 36:08and we're able to prove
• 36:09that these are all parallel
• 36:10clones. So independent genetic events
• 36:13developing,
• 36:14within this dysfunctional marrow. Here,
• 36:16these are individual mutations and
• 36:18columns, individual
• 36:20clones or cells,
• 36:21in rows.
• 36:22And so these are is
• 36:23the patient, for example, with
• 36:24nine different clones,
• 36:26all
• 36:27separate,
• 36:28genetically.
• 36:29When we look over
• 36:31time, most of these mutations
• 36:32are stable. They don't do
• 36:33anything,
• 36:34even the p fifty three
• 36:35mutations. And so having a
• 36:37p fifty three mutation doesn't
• 36:38tell you you're imminently
• 36:40transforming,
• 36:41and this will resonate if
• 36:42you've ever seen p fifty
• 36:43three clonal hematopoiesis.
• 36:46What does matter
• 36:47is all the leukemias
• 36:49had biallelic inactivation
• 36:51through LOH of various sorts
• 36:53or biallelic point mutations.
• 36:55And we can even take
• 36:56a patient who developed one
• 36:57of those leukemias
• 36:58and track back
• 37:00six years of their serial
• 37:01samples
• 37:02and identify,
• 37:04among these thirteen clones that
• 37:06they had, identify that moment
• 37:09five years before their transformation
• 37:10when they got their point
• 37:11one percent
• 37:13biallelic.
• 37:13They lost their heterozygosity,
• 37:15and that that then started
• 37:17to grow. This is on
• 37:18log scale. Started to just
• 37:20grow, grow,
• 37:21grow. All the other ones
• 37:22were stable. And and at
• 37:23the this last interval, it
• 37:25just blasted off. And so
• 37:27maybe that's a mechanism or
• 37:28pathway for rational surveillance.
• 37:30We can identify incipient,
• 37:33transformation
• 37:34by,
• 37:36identifying at risk clones within
• 37:38this c. We try to
• 37:40balance when to deploy transplant,
• 37:42because if any of you
• 37:43have ever transplanted someone,
• 37:46there are
• 37:47risks of commission,
• 37:50there if you,
• 37:51when you think about toxicities.
• 37:53And so this is just
• 37:54reiterating the model.
• 37:55Yeah. Six mutations
• 37:58repair the defect,
• 38:00but don't allow for transformation.
• 38:02P fifty three mutation monolemic,
• 38:05allow for increased growth, but
• 38:07not transformation,
• 38:08and then the biallelic hits,
• 38:10drive leukemia.
• 38:12So is this a generalizable
• 38:14model?
• 38:15Here's another patient.
• 38:17This is
• 38:18a was a kind of
• 38:19a weird
• 38:20one where we were doing
• 38:22some sequencing.
• 38:23We found a patient with
• 38:24MDS, seventy years old, no
• 38:25family history, had a p
• 38:27fifty three mutation and a
• 38:28variant of uncertain significance in
• 38:30TERT.
• 38:31Patient also had short ish
• 38:34telomeres.
• 38:36Couldn't be seventy years old
• 38:37with a no family history,
• 38:38but a germline
• 38:40telomere disease.
• 38:42And so we looked into
• 38:43this a little bit more.
• 38:44TERT is,
• 38:46has extremely high degree of,
• 38:49constraint,
• 38:50for missense substitutions or predict
• 38:53a loss of function mutations
• 38:54in in the genome. So
• 38:56it's a very tightly constrained
• 38:58gene.
• 38:59Here is among all genes
• 39:00and here is among even
• 39:01telomere associated genes.
• 39:03When it's mutated,
• 39:04the textbooks tell us that
• 39:05there's a mucocutaneous
• 39:07triad
• 39:07of,
• 39:09oral leukoplakia,
• 39:10nail dystrophy.
• 39:12There's also bone marrow failure
• 39:14and exceedingly short telomeres.
• 39:16And this is not this
• 39:17patient.
• 39:18So we looked in these,
• 39:20in MDS patients,
• 39:21sequenced all of the telomere
• 39:24related genes,
• 39:25identified all the variants, split
• 39:26them into common and rare
• 39:27with the hypothesis that rare
• 39:28variants
• 39:29might have snuck through evolution,
• 39:32and,
• 39:33be driving some telomere maintenance
• 39:35defect.
• 39:37So in MDS patients,
• 39:39the terp rare variance, terc
• 39:40and d k c one,
• 39:41were associated with short telomeres,
• 39:43relative to others.
• 39:45And,
• 39:46when we,
• 39:47looked at these candidate mutations,
• 39:49functionally,
• 39:50we cloned them all and
• 39:51tested them. Ninety percent of
• 39:53them or so were actually
• 39:54impaired,
• 39:55many of them severely impaired,
• 40:00suggesting that there's a germline,
• 40:03telomere
• 40:04dysfunction that's driving MDS biology.
• 40:06Does this matter for patients?
• 40:08Yes.
• 40:09TERT patients with TERT rare
• 40:10variance had reduced overall survival.
• 40:13This wasn't due to relapse.
• 40:14It was all due to
• 40:15toxicity.
• 40:16It was exactly what you
• 40:17would expect,
• 40:18noninfectious pulmonary disease,
• 40:20things like that that are
• 40:21associated with telomere disease. And
• 40:23when we look at even
• 40:25more resolution,
• 40:26it's those who get intensive
• 40:28conditioning, myeloblative conditioning, who really
• 40:30have that jump in in,
• 40:33NRM, and this is including
• 40:34the TERT and TURCs.
• 40:35But here, we're looking at
• 40:37more than fifty percent nonrelapse
• 40:38mortality with ablative conditioning with
• 40:40a TERT or TURC rare
• 40:42variant in MDS
• 40:44adults, not with TBDs. These
• 40:46are not known to have
• 40:47telomere disease.
• 40:48So we incorporated this into
• 40:49our standard panel,
• 40:51sequencing back in two thousand
• 40:52nineteen. So all patients with
• 40:54hematologic abnormalities
• 40:56at Dana Farber will get
• 40:57this test.
• 40:59And we've,
• 41:01I think identified since then
• 41:02about
• 41:03a hundred and nine patients
• 41:04with TERT rare variants, among
• 41:06patients with,
• 41:07largely myeloid malignancies.
• 41:10Few of them are pathogenic
• 41:12or likely pathogenic by ACGME
• 41:14criteria,
• 41:15and most of them are
• 41:16VUSs.
• 41:17And so most of them
• 41:18go in and,
• 41:20the clinician will say, okay.
• 41:22I don't know what this
• 41:23means,
• 41:25and just watch.
• 41:28So we measure the telomere
• 41:30length in all of these
• 41:30patients by flow fish, and
• 41:32sure enough, they have short
• 41:33telomeres.
• 41:34We measure the functional impact
• 41:36of all of these variants.
• 41:38Many of them
• 41:39have severely impaired,
• 41:41telomere extension,
• 41:43here in this kind of
• 41:43red pink area,
• 41:45intermediate
• 41:46here in white, or
• 41:48some of them were wild
• 41:49type as expected,
• 41:51in blue.
• 41:52This really raised the question
• 41:53of just
• 41:55how
• 41:55broad
• 41:56is this,
• 41:58is this observation.
• 42:00So
• 42:01how common are TIRT rare
• 42:03variants,
• 42:04and how much,
• 42:06explanatory
• 42:07power might they have for
• 42:09MDS and AML or even
• 42:11cancer more broadly.
• 42:14And so we hear this
• 42:15is just looking at, you
• 42:16know, UK Biobank, all of
• 42:18us, Nomad,
• 42:19showing the distribution of these
• 42:21rare variants or all variants.
• 42:24Most of them are only
• 42:25detected in one or two
• 42:27individuals.
• 42:28They're across all domains,
• 42:30but,
• 42:34and ninety six percent of
• 42:35them are VUSs. And so
• 42:37that's why we never hear
• 42:38about them, talk about them,
• 42:40think about them is because
• 42:41they're VUSs.
• 42:42And that's why no one
• 42:43associates
• 42:44TERT with,
• 42:46until recently with some of
• 42:47these, things.
• 42:48And so we decided to
• 42:50I think we identified something
• 42:51with lately. I think we're
• 42:52up to seventeen hundred
• 42:54tert rare variants,
• 42:55across all human
• 42:58population datasets.
• 42:59So we've cloned them all
• 43:01and developed a arrayed strategy
• 43:03for functional testing,
• 43:06where we do five biological
• 43:07replicates for all of them,
• 43:10express them inducibly in cells,
• 43:12and then measure telomere length.
• 43:14This is just validation using
• 43:16standard,
• 43:17kind of southern blot,
• 43:21just in a just in
• 43:22some.
• 43:23I'm not I'm not evil.
• 43:27QPCR is really the the
• 43:29scalable method by by which
• 43:31we do that. And so
• 43:32here's validation. Is it these
• 43:33are
• 43:35bonafide,
• 43:36telomere disease in the telomerase
• 43:38database. Here's how they perform.
• 43:40And so here are the
• 43:41the controls, and here are
• 43:43the,
• 43:44database
• 43:45samples.
• 43:46We actually do better than
• 43:47the standard trap assay,
• 43:49in part because we can
• 43:50identify,
• 43:51enzymatically,
• 43:54capable,
• 43:56TERT that has defects in
• 43:57localization or other mechanisms of
• 43:59of, function. And so many
• 44:02variants in TERT
• 44:05are shown to be wild
• 44:06type in the standard assay
• 44:07even when they don't work
• 44:09in cells because they can't
• 44:10get to the right place.
• 44:12And so,
• 44:14here is just,
• 44:15this was a data grab
• 44:16from a while back,
• 44:18looking at domain specificity of
• 44:20these variants when we look
• 44:22by,
• 44:24oops, when we look across,
• 44:26the VOSs.
• 44:28Linker, this is an unstructured
• 44:29linker. It doesn't do anything,
• 44:31and consistent with that, they're
• 44:32all wild type.
• 44:34And then there's tons of
• 44:36dysfunctional
• 44:37TERT in these rare variants
• 44:38or ultra rare variants.
• 44:40And,
• 44:42and here, I'm just showing
• 44:43you,
• 44:44the tools that we use
• 44:45clinically to to deconvolute VOSs
• 44:48are really
• 44:50poor. So these are our
• 44:51best current like, the current
• 44:53best deep learning models for
• 44:55variant effect prediction.
• 44:57Eve, uses,
• 44:59evolutionary conservation,
• 45:01to predict benign, uncertain, or
• 45:03pathogenic based on
• 45:05scores.
• 45:06Alpha missense uses
• 45:08alpha fold,
• 45:09predicted structures to identify or
• 45:11to categorize.
• 45:13What you can see here
• 45:14is
• 45:15these are
• 45:16the benign,
• 45:18intermediate, or pathogenic
• 45:20based on each of these
• 45:21prediction models. And then on
• 45:22the y axis is their
• 45:23functional score.
• 45:25And,
• 45:26you can see that the
• 45:27pathogenic gets pretty specific.
• 45:31Intermediate is
• 45:32across the board. And then
• 45:33the benign predictions, this is
• 45:35the real fault of these
• 45:36prediction models.
• 45:38Many of them are actually
• 45:39functionally impaired, including severely impaired
• 45:42or dead in the assay.
• 45:44And so,
• 45:46we kind of dig into
• 45:48this and say, why might
• 45:49this be? So I'm focusing
• 45:51on this ten domain,
• 45:52which is at the n
• 45:53terminal domain of TERT. It's
• 45:55in the first hundred and
• 45:56eighty amino acids. It's important
• 45:58for
• 45:58binding to,
• 46:00to TPP one, which is
• 46:02how TERT finds the right
• 46:04spot
• 46:05on the end of DNA
• 46:07to to act to extend
• 46:09telomeres.
• 46:10And I'm just showing you
• 46:11here, this is missense tolerance
• 46:13ratio. So this is kind
• 46:14of areas of the,
• 46:16that are the least tolerant
• 46:19to missense substitutions in in
• 46:21evolution.
• 46:22And you can see that
• 46:23these severely impaired variants
• 46:25are really centered on that
• 46:27intolerant
• 46:28area,
• 46:30and that more than half
• 46:31of the variants are dead.
• 46:33These are just regular people
• 46:35walking around. Half of them
• 46:37have dead telomerase,
• 46:38because of this defect.
• 46:40And why don't these things
• 46:41work,
• 46:42in the models?
• 46:44It's because,
• 46:45I think,
• 46:46because a ten domain mediates
• 46:49intermolecular
• 46:50interactions and intramolecular
• 46:52interactions that are poorly predicted
• 46:53by structure.
• 46:54And so those are,
• 46:59ten,
• 47:00domain TPP one interactions,
• 47:02and interactions,
• 47:04with,
• 47:05the the template, the TURC,
• 47:07the RNA template,
• 47:09among others.
• 47:10And so
• 47:12digging into it a little
• 47:13bit more, we've taken a
• 47:15look now at multiolelic residues,
• 47:19that have divergent,
• 47:22functional effects. And so here
• 47:24are just three examples,
• 47:27c seventy six, g one
• 47:28thirty five, and r ten
• 47:29eighty six, where one variant
• 47:31is dead and the other
• 47:33is
• 47:34preserved a wild type, saying
• 47:35maybe this will afford us
• 47:37some insight into
• 47:38where the models fail and
• 47:40why these variants have this,
• 47:43this these effects. And so
• 47:44here are these variants
• 47:46according to their alphamis since
• 47:47in Eve,
• 47:49most of them benign. And
• 47:50so this is a point
• 47:52of failure for the models.
• 47:54So that's why we're digging
• 47:55into these. We use molecular
• 47:57dynamic simulation
• 47:58to,
• 47:59which is essentially like,
• 48:01a computational approach that uses
• 48:03basic rules of physics and
• 48:06inter and molecular,
• 48:08function
• 48:09to,
• 48:11not take static pictures like
• 48:12you would see in a
• 48:13crystal structure at four,
• 48:15but predict how molecules
• 48:17interact
• 48:18over time iteratively.
• 48:21And so
• 48:22what we can see here
• 48:23is that if we look
• 48:24at this,
• 48:25this is the g one
• 48:26thirty five,
• 48:28result.
• 48:29The g one thirty five
• 48:32r here,
• 48:34it actually forms a salt
• 48:35bridge without altering the confirmation.
• 48:38And as a result, it
• 48:40actually has slightly better contacts,
• 48:42with TPP one. So it
• 48:44actually performs well, if not
• 48:46better, than wild type. Whereas
• 48:47this g one thirty five
• 48:49e,
• 48:51incurs electrostatic
• 48:52repulsion
• 48:53because there it's a rated
• 48:55in a,
• 48:57like, a acidic patch of
• 48:58TPP one. And so it
• 49:00drives
• 49:01decreased contact and breaks that
• 49:03recruitment interaction.
• 49:05And so we're now
• 49:07using this approach to build
• 49:09a new deep learning model
• 49:11for variant effect prediction that
• 49:12incorporates these structure function,
• 49:15findings that we have, and
• 49:17I'll skip that for now.
• 49:19And so
• 49:20moving to the last little
• 49:21bit,
• 49:23let's connect it back to
• 49:24disease initiation. So this is
• 49:25that's setting the stage of,
• 49:27that was a rabbit hole.
• 49:28Let's just call it that.
• 49:29Call it what it is.
• 49:31It was a rabbit hole
• 49:32for us because
• 49:33we were confronted with VUSs.
• 49:36And if any of you
• 49:36ever have to deal with
• 49:38VUSs, you know that they
• 49:39should
• 49:40inspire,
• 49:41loathing,
• 49:42and frustration,
• 49:45because
• 49:46we know that they either
• 49:47are one thing or another.
• 49:49And some and when they
• 49:50are
• 49:52the when they're pathogenic and,
• 49:55you don't know that, or
• 49:56you only know it retrospectively,
• 49:58the implications for patients can
• 49:59be severe.
• 50:01So these are patients, if
• 50:02you take them to a
• 50:03myeloblade of aloe, you're incurring,
• 50:05as I showed you before,
• 50:06a sixty percent chance of
• 50:08nonrelapse mortality.
• 50:09And so these are this
• 50:10is meaningful. And so that
• 50:12was that was the rabbit
• 50:13hole.
• 50:15Hopefully, the postdocs
• 50:17like the rabbit hole.
• 50:19But here's getting back to
• 50:20disease initiation,
• 50:22if we say aging is
• 50:23associated with decreased fitness,
• 50:25there are multiple ways that
• 50:26aging could do that, one
• 50:27of which could be the
• 50:29cumulative effects of
• 50:31replication.
• 50:32And so stem cells undergo
• 50:34telomere attrition as they
• 50:36replenish the blood. And so
• 50:38maybe this is a fitness
• 50:40constraint
• 50:40that evolves over age and
• 50:42could explain the age associated
• 50:43risk of leukemia
• 50:44in a subset of the
• 50:45population. And so here's what
• 50:47the fitness would look like,
• 50:48and here's, you know, if
• 50:49you have that fitness defect,
• 50:51you could, select out clones
• 50:53there.
• 50:54And so sure enough,
• 50:56clonal hematopoiesis in patients with
• 50:58in our telomere disease program,
• 51:01they have a very different
• 51:03spectrum
• 51:04of CH than healthy donors
• 51:06who are DNMT three and
• 51:07tattoo.
• 51:07We see lots of PPM1D,
• 51:09p fifty three, U2F1, s
• 51:11thirty four.
• 51:12When we do clonal decomposition
• 51:14with single cell, these are
• 51:16all, again, independent clones,
• 51:18growing out.
• 51:19And then those that transform,
• 51:21the genetics are different than
• 51:23the baseline CH.
• 51:24It's p fifty three
• 51:27and u two f one
• 51:28s thirty four f,
• 51:30which drive transformation
• 51:31in patients with telomere disease,
• 51:33and there's no PPM1 d.
• 51:35And so maybe this is
• 51:36an EI six,
• 51:38p fifty three analogy,
• 51:40for,
• 51:42t for it to like,
• 51:43Wish Walkman.
• 51:44So what is PPM1D?
• 51:45It's basically the the negative
• 51:47regulator of the entire DNA
• 51:49damage response.
• 51:50It's a phosphatase,
• 51:52that,
• 51:54that regulates ATM,
• 51:56check one, ATR, p fifty
• 51:57three, check two, MDM two,
• 51:59gametes two ax, everything. It's
• 52:01the return to normal,
• 52:03that's required for,
• 52:05regulation of the DDR.
• 52:07The mutations
• 52:08lop off a degron
• 52:10resulting in a highly stabilized
• 52:11phosphatase that's hyperactive,
• 52:13and so this quiets down,
• 52:15the DDR.
• 52:17That's what happens in with
• 52:18telomere attrition.
• 52:20So with every replication,
• 52:24before s phase, the telomere
• 52:25is capped. It's kind of
• 52:26looped over itself.
• 52:27It has to uncap.
• 52:30In
• 52:30the process, it becomes deprotected,
• 52:34and is at risk for
• 52:35being recognized as a double
• 52:36stranded break,
• 52:38or DNA damage.
• 52:40Telomerase sits on it,
• 52:42extends
• 52:43a few times,
• 52:44then it recaps, and,
• 52:46and then we're good.
• 52:48If, in the setting of
• 52:49attrition,
• 52:50eventually,
• 52:51that cap it it has
• 52:52a hard time actually recapping,
• 52:54and that state of, vulnerability
• 52:57is extended,
• 52:58and DDR is activated above
• 53:00threshold.
• 53:01And so you get a
• 53:02ATM dependent check two p
• 53:04p three,
• 53:05senescence program
• 53:07that activates,
• 53:08g two and g one
• 53:09s checkpoints and and drive
• 53:11senescence.
• 53:12So you could imagine that
• 53:13mutations in ATM check two,
• 53:16PPM1D or p fifty three
• 53:17could in different ways attenuate
• 53:19these steps.
• 53:22So we have developed models,
• 53:23and I'll just I'm not
• 53:24gonna go into this, but,
• 53:26where we engineer,
• 53:29natural attrition,
• 53:31that results in a stereotyped,
• 53:33deprotection,
• 53:35of telomeres. They activate the
• 53:37that entire program, and this
• 53:38is CDKM one a. And
• 53:40so they go through this
• 53:41this process.
• 53:43When they reach that point,
• 53:44they do the right things
• 53:45when it comes to cell
• 53:46cycle arrest,
• 53:48or checkpoint activation. And we
• 53:50use that as a way
• 53:51to then,
• 53:54model the effects of all
• 53:56these different mutations
• 53:57on the naturally occurring,
• 54:00telomere deprotection.
• 54:02And I,
• 54:03didn't correct this slide, but,
• 54:06essentially, when we compare PPM1D
• 54:07and p fifty three, p
• 54:09fifty three mutations
• 54:10allow these cells to escape,
• 54:13senescence, and so they just
• 54:14keep growing when they should
• 54:16stop. And PPM1D just extends
• 54:18it by a a passage
• 54:19or and a half.
• 54:22And so it delays the
• 54:24activation of,
• 54:26p twenty one,
• 54:29but is overwhelable
• 54:31by,
• 54:32continued DDR signaling.
• 54:35P fifty three actually activates
• 54:36a whole second,
• 54:39kind of contingency
• 54:41pathway
• 54:42that I won't get into
• 54:44now. But we're digging into
• 54:46now this connection between,
• 54:48deep protection and conal selection.
• 54:51And so how broad is
• 54:52this issue?
• 54:53This is when we look
• 54:54at the UK Biobank,
• 54:56we see many genes where
• 54:58their rare variants impact telomere
• 55:00length either to shorten it
• 55:02or to longer,
• 55:03lengthen it.
• 55:05And so this as a
• 55:06paradigm of aging and cancer
• 55:08risk is something that we're
• 55:09broadly
• 55:10engaged in now. And this
• 55:12cuts along
• 55:13which whether these are affect
• 55:15the,
• 55:16the c strand of the
• 55:17telomere or the g strand.
• 55:19And this is essentially,
• 55:20over here. There's a g
• 55:22risk strand, which,
• 55:24which is where that telomerase
• 55:25stuff happens. And then this,
• 55:28CRIST strand, when there are
• 55:29mutations,
• 55:30in genes that regulate that,
• 55:31they actually cause aberrant lengthening.
• 55:33And so this is, an
• 55:35area of of future
• 55:37ex of, ongoing exploration.
• 55:39And so we're asking questions
• 55:40like, is
• 55:41the,
• 55:43age associated,
• 55:45telomere attrition really a generalizable
• 55:47mechanism
• 55:48by which,
• 55:50individuals have MDS risk?
• 55:54Do patients on the with
• 55:56telomere disease kind of accelerate
• 55:58that process?
• 55:59Is it different?
• 56:01And then,
• 56:02as a teaser, I'll show
• 56:04you we've now used,
• 56:06developed and deployed,
• 56:08ultra sensitive duplex sequencing
• 56:10panel that's scalable
• 56:12now,
• 56:13so we can do studies
• 56:14in
• 56:15I think we've done probably
• 56:17twelve to thirteen thousand samples
• 56:19now. But this is ultra
• 56:20sensitive. This is a level
• 56:21of detection down,
• 56:23below one point one percent
• 56:25parent to little fraction.
• 56:26You can see that mutations
• 56:28accumulate
• 56:29with age.
• 56:30We've done the negative controls
• 56:31that the infants are dead
• 56:33negative. But what you can
• 56:34see here is that here's
• 56:35the CH you know about
• 56:36where you where we focus
• 56:38on DNMT three eight tattoo
• 56:39ASX zero one. This is
• 56:40relatively young,
• 56:41and then there's this whole
• 56:42wave of DDR CH,
• 56:44that happens later,
• 56:46really raising the possibility if
• 56:47that's what happens.
• 56:49And so this is, I
• 56:50think, how CH works over
• 56:52time
• 56:52with evolving fitness.
• 56:54And so when we sequence,
• 56:56we we take a slice
• 56:57at any one of these
• 56:58time points, and the CH
• 56:59that we see is the
• 57:00CH,
• 57:02that fitness got us.
• 57:04And so,
• 57:06that's, what I have for
• 57:07you today.
• 57:08Chris Riley is, really the
• 57:10partner who drove this when
• 57:11he was a postdoc in
• 57:12my lab and is now
• 57:14running a large
• 57:16telomere disease multidisciplinary
• 57:17program. So if you ever
• 57:18have variants,
• 57:20call him or email him.
• 57:22He will give you
• 57:23a comprehensive evaluation
• 57:25and refer patients to him.
• 57:26He's, that's all he does.
• 57:28And so lots of collaborators,
• 57:29and, thanks for your time.
• 57:37And I recognize I went
• 57:39to my time, so I'm
• 57:40happy to talk with anyone
• 57:41at any at any,
• 57:43point.
• 57:45Yes.
• 57:46Okay. Thank you.
• 57:49Found it interesting that you,
• 57:51talked about schwaltz and diamond
• 57:52first and it's not a
• 57:53whole,
• 57:55for white column, and then
• 57:56you went into intelligence. But
• 57:58there's a report here in
• 57:59the back showing that guacamole
• 58:01diamond protein might play a
• 58:02role in telomeres too. I
• 58:04wonder if you find that
• 58:06that's a possibility
• 58:08compatible with what
• 58:09you tell them or I
• 58:11everything that we see from
• 58:13the,
• 58:14so I view,
• 58:15this sort of sequencing approach
• 58:17as a as a CRISPR
• 58:18screen, essentially, or and so
• 58:21the
• 58:21clonematopoiesis
• 58:22that we see growing out
• 58:23in Schwackman Diamond syndrome
• 58:25is all,
• 58:28related to things so p
• 58:30fifty three and then everything
• 58:31else is about fixing ribosomal
• 58:34p fifty three activation.
• 58:36So translational stress. So we
• 58:38see mutations in,
• 58:40dead box helicase
• 58:42proteins
• 58:43in RPL five, RPL twenty
• 58:44two,
• 58:46all ribosome based. And I
• 58:48don't see ATM check two
• 58:50PPM1D,
• 58:51which are the now the
• 58:53signature of telomere stress.
• 58:55So I would say
• 58:56the humans tell us that
• 58:57that's not true. Like, that's
• 58:59not that driving selection pressure.
• 59:01I should be more accurate.
• 59:03Does that make sense? Like
• 59:04yeah.
• 59:05I like following the the
• 59:07the humans really just tell
• 59:09you what does happen,
• 59:12through selection. I think that's
• 59:13the
• 59:17Yeah.
• 59:18Great talk, Coleman, as usual.
• 59:20So for the t p
• 59:21fifty three, in the Vyxeos
• 59:23study,
• 59:24like you, I was surprised
• 59:25with the twenty five percent,
• 59:27induction or sixty day mortality.
• 59:30So I understand this is
• 59:31with both drugs, the same
• 59:32twenty five percent? Or Yeah.
• 59:34It was it was, I
• 59:35didn't say that, but the
• 59:37early mortality was the same
• 59:39across molecular subtypes irrespective
• 59:41of drug.
• 59:43So is your sense that
• 59:44those patients,
• 59:45are dying sooner? Because, you
• 59:47know,
• 59:48I I guess the other
• 59:50part of the question is
• 59:51that this study was sixty
• 59:52years and and older. So
• 59:54the the first question is,
• 59:55do you think this applies
• 59:56even for younger patients with
• 59:58TB fifty three? And the
• 60:00main question is, do you
• 01:00:01think that this is a
• 01:00:02optimistic clinician. I I love
• 01:00:04you, Amor.
• 01:00:05Yeah. And do do you
• 01:00:06think this is,
• 01:00:09reflection of primary induction failure
• 01:00:11and fungal infection and Yeah.
• 01:00:12It's gonna be there. Or
• 01:00:13is it, like, some excessive
• 01:00:15bartle,
• 01:00:16chemo related toxicity that we
• 01:00:18just don't I think it's
• 01:00:19a composite. I think that's,
• 01:00:21I think it's a composite.
• 01:00:22I will say that there's
• 01:00:23a strong association be and
• 01:00:25I'm I think I'm not
• 01:00:26gonna get into this, but,
• 01:00:28between albumin less than three
• 01:00:31and early mortality,
• 01:00:33generally, but then in this
• 01:00:34trial for sure.
• 01:00:36And so I think what
• 01:00:37does that say about a
• 01:00:38patient?
• 01:00:39You tell me, but about
• 01:00:40the that tells us maybe
• 01:00:42it's not just about induction,
• 01:00:44failure, but there's some,
• 01:00:47something about the host
• 01:00:48there that
• 01:00:49that might influence, but we
• 01:00:51don't have the data to
• 01:00:52to deconvolute that. Thanks.
• 01:00:57Thank you for that. Great
• 01:00:58talk. You know, you did
• 01:01:00show the TURK mutations.
• 01:01:02The nonelapse mortality is fifty
• 01:01:04percent.
• 01:01:05I would lose my job.
• 01:01:07What's driving it?
• 01:01:10It
• 01:01:11so
• 01:01:12cause of death in transplant
• 01:01:14is is tough
• 01:01:15to ascertain oftentimes,
• 01:01:17but there is a markedly
• 01:01:18increased risk of,
• 01:01:20noninfectious pulmonary disease
• 01:01:22of the,
• 01:01:24there's also
• 01:01:25a more of a consequence
• 01:01:26of acute gut GVHD.
• 01:01:28And so there's no difference
• 01:01:30in GVHD, the cumulative incidence
• 01:01:32of GVHD,
• 01:01:33but the consequences of severe
• 01:01:34gut GVHD are are more
• 01:01:36in patients with short telomeres.
• 01:01:38You can imagine because they
• 01:01:40have less regenerative potential after
• 01:01:42injury of the gut mucosa.
• 01:01:44We have mouse models actually
• 01:01:46to show that as well.
• 01:01:47If you induce GVHD in
• 01:01:49a mouse,
• 01:01:50they
• 01:01:51with with short telomeres, they
• 01:01:53die of gut failure.
• 01:01:55So that I think those
• 01:01:57those types of things are
• 01:01:58what they die
• 01:02:00Right. This is, this is
• 01:02:01definitely kind of a shock.
• 01:02:02You know, there are problems
• 01:02:05for a transplant that early,
• 01:02:07but then you show on
• 01:02:08the other side of the
• 01:02:09story.
• 01:02:10Well, yeah, I guess, it's
• 01:02:11mode of transplant. So I
• 01:02:12would take a,
• 01:02:14would not do an ablative
• 01:02:15transplant in these patients and
• 01:02:17maybe reduce intensity or,
• 01:02:20other
• 01:02:21like, a a toxicity toxicity
• 01:02:23sparing strategy would be better.
• 01:02:29I have a practical question.
• 01:02:31Sure.
• 01:02:32To detect short telomeres,
• 01:02:36how
• 01:02:37confident can we be from
• 01:02:38our flow FISH
• 01:02:39result,
• 01:02:41and is there a commercial
• 01:02:43PCR assay that would be
• 01:02:44appropriate for use? So the
• 01:02:46the best assay that I
• 01:02:48know of, that, that we
• 01:02:49use is the repeat,
• 01:02:52diagnostics,
• 01:02:54flow fish telomere length testing.
• 01:02:57What I will say is
• 01:02:58that,
• 01:03:00as you age,
• 01:03:01the normal
• 01:03:03physiologic telomere length,
• 01:03:05approaches pathologic telomere length. And
• 01:03:08so the test in terms
• 01:03:09of age adjusted adjusted percentile
• 01:03:12becomes meaningless,
• 01:03:14when you're non pediatric because
• 01:03:16all the kids have very
• 01:03:17long telomeres, and so there's
• 01:03:18they separate out normal and
• 01:03:20pathologic.
• 01:03:21But as you go across
• 01:03:22time, pathologic remains the same
• 01:03:24because the cell doesn't care
• 01:03:25whether you're five years old
• 01:03:26or eighty years old. Right.
• 01:03:28Short telomere, short telomere.
• 01:03:30But then aging
• 01:03:31brings it down
• 01:03:33ubiquitously.
• 01:03:34So it's not a very
• 01:03:35informative test. So it's not
• 01:03:37sensitive.
• 01:03:38So you can definitely have
• 01:03:40if you take a genotype
• 01:03:41first approach,
• 01:03:43you can definitely have people
• 01:03:44with dysfunctional telomere maintenance who
• 01:03:46have normal appearing telomeres,
• 01:03:48although they tend to be
• 01:03:49less than fiftieth percentile
• 01:03:51by age
• 01:03:53or,
• 01:03:54less than, like, four and
• 01:03:56a half kb or something
• 01:03:57like that. But remember, Flowfish
• 01:03:59is also population level,
• 01:04:01like cell population,
• 01:04:03or
• 01:04:04telomere population. It's single cell
• 01:04:07flow. Mhmm. But
• 01:04:09what matters is how many
• 01:04:11short telomeres you have within
• 01:04:12a cell, and there's a
• 01:04:13lot of heterogeneity.
• 01:04:15We can get into that
• 01:04:16later. But so Yeah.
• 01:04:19I think sounds like a
• 01:04:20genetic Right now, we're taking
• 01:04:22a genotype first approach,
• 01:04:24as the initial.
• 01:04:30K. Thanks, everyone.